Blind source separation using analysis sparse constraint

Abstract

A novel algorithm based on the analysis sparse constraint of the source over an adaptive dictionary is proposed to solve the blind source separation problem. In the algorithm, the dictionary for each source is adaptively learned from the corresponding source, which is estimated from the mixtures. Moreover, then the analysis sparse representation of the source can be obtained with the learning dictionary. The representation of the source is the constraint that can be employed to extract the source from the mixtures. Experimental results demonstrate that the proposed method improves separation performance.

Speaker Recognition Based on Long-Term Acoustic Features With Analysis Sparse Representation

“…x, where f ∈ R P×1 is the sparse representation of x. More recently, the analysis sparse model has been drawing increasing attention due to its application in image denoising [22], source separation [23], [24], image encryption [25], [26] and image classification [27], [28]. In [27], based on the analysis sparse model, the sparse representations of an image can be learned and used as the features of the image for training a support vector machine to resolve the problems of the image classification.…”